Overview
Explore a conference talk from USENIX PEPR '20 that delves into LinkedIn's implementation of a differentially private data analytics API. Learn how this system protects member data while providing valuable audience engagement insights for marketing analytics. Discover the differentially private algorithms and privacy safeguards used to integrate with real-time data analytics platforms, particularly the open-source Pinot system. Understand the user-level privacy guarantees and the role of the budget management service in enforcing strict differential privacy budgets. Gain insights into how LinkedIn balances utility and privacy in their data analytics approach, incorporating the latest research in differential privacy into a practical product solution.
Syllabus
Intro
Our Mission
Differential Privacy
Differentially Private
Models of Differential Privacy
Audience Engagement API
What is a differencing attack
Existing infrastructure
Privacy system
Sensitivity
Summary
Known vs Unknown Domain
Algorithms
Overall Privacy
Taught by
USENIX